research

Delegate the question to a background agent that reads official docs, source code, and specs directly. It follows each claim back to the source that owns it and writes a cited Markdown report to your repo.

Primary sources are official documentation, source code, specifications, and first-party APIs. Secondary write-ups like blog posts or tutorials are avoided because claims must trace back to the source that owns them.

The report is saved as a single Markdown file where the repo already keeps such notes, matching existing conventions. If no convention exists, it is placed somewhere sensible and the location is reported.

Yes, the investigation runs as a background agent, so you can continue other work while it reads sources and compiles the findings into the Markdown file.

The agent only reports what primary sources state, so undocumented behavior or questions requiring live experimentation may not be fully answered. Findings quality depends on the availability and clarity of official sources.